It's interesting to see how well ChatGPT works and where its (current) limitations are, so I chatted with it about some of our SQL challenges. I found it a fun read!
At the end of the two-weeks primer course on SQL, Tableau and Python, everybody builds a simple text-based Hangman (click on the image for my take on it):
In this case study, the company Eniac wants to expand its business to Brazil and evaluates the potential after-sales fulfillment partner Magist for its suitability.
As second basic data-handling system after SQL, we were introduced to the Python pandas library. Read about our challenges here!
In this quite intense two-weeks chapter, we were diving deep into a sales database with severe problems and learned how to still extract useful conclusions from it. Take a look!
In this one-week chapter, we learned a lot about the foundations of inferential statistics and deciding whether the outcome of UI experiments have statistical significance. Because the methods are applicable in a much larger class of problems, I found the material very helpful.
In this two-week project, we learned and exercised ETL data-engineering skills, i.e. extracting, transforming and loading data into storage for comprehensive analysis. We scraped the web, used public APIs, transformed and augmented the data and stored it in an SQL database. The finished ETL process was then wrapped into a Google cloud function for automatic execution and I even went further to produce automatically updated reports on the data.
One of the deliverables was a blog post which I wrote on dev.to.
In this one-week project we learned about high dimensional distances, scaling, PCA, k-Means, inertia elbow and silhouette score and the Spotify API.
My special treat was to apply harmony theory to order songs by harmonic distance.
Two weeks were devoted and crammed with insights into supervised machine-learning. We learned about
- training data preparation
- classification, regression
- prediction metrics
- decision trees, gradient boosted random forests
- linear and logistic regression
- support vector classifiers
- one-hot and ordinal encoding
- parameter optimization and cross-validation
and even pickling data and creating classifiers as web-apps with streamlit! Our model data-sets were selling prices of houses 🏰 and poisonous vs. edible mushrooms 🍄.
This week took us to learn about different ways to extract movie recommendations for the fictitious WBSFLIX online DVD rental shop from previous movie ratings. Read all about it here and check out the recommendation app!
I actually do like SQL a lot, so this one-week reinforcement on advances SQL topics was actually good fun!
coming soon